Jordan Madison
Research Engineer building post-training, evaluation, and infrastructure for open-weight language models.
Interested in research engineering, ML engineering, and MTS roles.
I'm a Georgia Tech computer scientist working at the intersection of machine learning, infrastructure, and security. I still play Rocket League and spend more time than I should wondering how many bits it takes to reach the optimum of a loss curve.
At Georgia Tech's Reasoning and Learning Group, I build the environments and tooling behind CreSTAL, an in-silico materials-editing project. I first used Qwen3-0.6B to validate the end-to-end harness, then expanded the system to support the reported study across four open-weight models: Qwen3-4B, Minitron-4B, Llama 3.2 3B, and Phi-4 Mini.
My work spans executable RL environments, post-training pipelines, HPC orchestration, and the debugging required to make these systems reliable. I also have a broader background in applied machine learning, cybersecurity, privacy, and malware reverse engineering. I like turning ambiguous technical failures into measurable experiments and reusable infrastructure, especially where machine learning, infrastructure, and security meet.
Selected work
-
CreSTAL · Post-training research
An in-silico materials-targeting RL environment for studying post-training and reward design. Across four reported models, held-out smooth property utility increased after RL while strict-hit counts moved in different directions.
Read the case study → -
Atlas · Research infrastructure
A job orchestration layer for post-training research that turns reusable templates into repeatable SLURM training, evaluation, and data-generation workflows.
View the project → -
Cloudflare Radar · 0-to-1 product
Led 0-to-1 product work that turned global network telemetry into a public internet intelligence platform shipped in 12 weeks.
View experience →
Recent writing
-
An application of self-improving harnesses to limit order placement.
#design , #finance , #investment , #agent , #meta
-
A breakdown of an RL environment I designed after misreading a take-home assignment. The design is built around my reverse-engineering background and I liked it enough to write it up here.
-
A LoRA debugging story about four embeddings that never got to learn anything. If you add tokens and freeze the base model, their embedding and LM-head rows still need to train.
#post-training , #sft , #lora , #tokenizers , #ai-assisted
-
Reading GPU telemetry on a colocated GRPO run: periodic utilization dips, memory-access spikes, and a bottleneck story of mine that didn't survive contact with the code.
#rl , #post-training , #grpo , #vllm , #ai-assisted
-
A debugging story about special tokens, GRPO, and the skip_special_tokens flags scattered across your RL stack. When your SFT'd model nails the format in eval, then falls apart the moment it enters the GRPO rollout.
#rl , #post-training , #grpo , #tokenizers , #ai-assisted